Measuring AI adoption without telemetry: a 400-person firm's survey-plus-Claude approach
Expert-Birthday7928 · reddit · 2026-10-17
A 400-person firm working across 100+ client projects can't use telemetry (commits, tickets) to measure AI adoption, so it surveys the human side instead. Findings: self-reported usage runs high but trends are meaningful; reported blockers like unclear client rules are invisible to telemetry; running results through Claude to ask 'which teams are stuck and why' beats dashboards. The author, who is building a tool around this, asks how far self-reporting deviates from real usage data.
More from Companies & People
- How product leaders can harness AI early adopters without distracting the whole team — lennysan · 2026-10-17
- Anthropic bets voluntary scrutiny sets AI oversight terms before regulators act — VectorWireAI · 2026-10-17
- Google CEO Pichai calls Elon Musk's ability to 'will future technologies into existence' unparalleled — XFreeze · 2026-10-17
- One user made 24,000 Copilot requests in 24h: how firms audit 580 AI seats nobody tracks — alex_verem · 2026-10-17
- Ion Stoica at Stanford: humans specify and verify, models design and generate — RajaXg · 2026-10-17
- Arcee AI on a 'generational' hiring run, snapping up top talent — TheZachMueller · 2026-10-17